How to handle the undocumented device config spec?

[Wikilicious]

A major factor in my decision to start this project was the GQ Geiger Counter Communication Protocol documentation! https://www.gqelectronicsllc.com/download/GQ-RFC1801.txt & https://www.gqelectronicsllc.com/download/GQ-RFC1201.txt

I slowly expanded support for the commands, and it now feels like the final frontier is the device configuration - a 256 or 512 bytes data dump.

The issue is... that's a lot of un-documented bytes!

There are forum posts on https://www.gqelectronicsllc.com/forum/forum.asp?FORUM_ID=14 that detail what each byte means... however; it's more of a reactive-than-proactive support e.g. https://www.gqelectronicsllc.com/forum/topic.asp?TOPIC_ID=10394 & https://www.gqelectronicsllc.com/forum/topic.asp?TOPIC_ID=10437

It seems the config spec (other than config size) is unrelated to the device or communication spec.

I can handle mapping rev-version (rev version is included in the getver cmd) and re-architect config to load the correct revision for a device... but what I am stuck on is... do I even support something with no specs and most importantly no revision mapping?
As seen in the link above, a user will inevitably submit an issue that pygmc isn't working as expected and the root-cause will be that GQ made a new config.

[Wikilicious]

Pathfinding... without an exact spec to support... it feels like a lost battle.
I can put in a lot of time and effort yet they can willy-nilly change "specs" without any documentation; completely canceling out my efforts.